The Effect of Race and Chronic Obstructive Pulmonary Disease on Long-Term Survival after Coronary Artery Bypass Grafting

CitationEfird, Jimmy T., Wesley T. O’Neal, Curtis A. Anderson, Jason B. O’Neal, Linda C. Kindell, T. Bruce Ferguson, W. Randolph Chitwood, and Alan P. Kypson. 2013. “The Effect of Race and Chronic Obstructive Pulmonary Disease on Long-Term Survival after Coronary Artery Bypass Grafting.” Frontiers in Public Health 1 (1): 4. doi:10.3389/fpubh.2013.00004. http://dx.doi.org/10.3389/fpubh.2013.00004.

Published Versiondoi:10.3389/fpubh.2013.00004

Permanent linkhttp://nrs.harvard.edu/urn-3:HUL.InstRepos:11877144

Terms of UseThis article was downloaded from Harvard University’s DASH repository, and is made available under the terms and conditions applicable to Other Posted Material, as set forth at http://nrs.harvard.edu/urn-3:HUL.InstRepos:dash.current.terms-of-use#LAA

Share Your StoryThe Harvard community has made this article openly available.Please share how this access benefits you. Submit a story .

Accessibility

PUBLIC HEALTHORIGINAL RESEARCH ARTICLE

published: 15 April 2013doi: 10.3389/fpubh.2013.00004

The effect of race and chronic obstructive pulmonarydisease on long-term survival after coronary arterybypass graftingJimmyT. Efird 1,2*,WesleyT. O’Neal 3, Curtis A. Anderson1, Jason B. O’Neal 4, Linda C. Kindell 1,T. Bruce Ferguson1,W. Randolph Chitwood 1 and Alan P. Kypson1

1 Department of Cardiovascular Sciences, East Carolina Heart Institute, Brody School of Medicine, East Carolina University, Greenville, NC, USA2 Center for Health Disparities Research, Brody School of Medicine, East Carolina University, Greenville, NC, USA3 Department of Internal Medicine, Wake Forest University School of Medicine, Winston-Salem, NC, USA4 Department of Anesthesia, Critical Care and Pain Medicine, Beth Israel Deaconess Medical Center, Harvard Medical School, Boston, MA, USA

Edited by:Fabrizio Stracci, University of Perugia,Italy

Reviewed by:Fabrizio Stracci, University of Perugia,ItalySetor K. Kunutsor, University ofCambridge, UK

*Correspondence:Jimmy T. Efird, Director, Epidemiologyand Outcomes Research Unit, EastCarolina Heart Institute, Brody Schoolof Medicine, East Carolina University,115 Heart Drive, Greenville, NC27834, USA.e-mail: jimmy.efird@stanfordalumni.org

Background: Chronic obstructive pulmonary disease (COPD) is a known predictor ofdecreased long-term survival after coronary artery bypass grafting (CABG). Differencesin survival by race have not been examined.

Methods: A retrospective cohort study was conducted of CABG patients between 2002and 2011. Long-term survival was compared in patients with and without COPD and strat-ified by race. Hazard ratios (HR) and 95% confidence intervals (CI) were computed usinga Cox regression model.

Results: A total of 984 (20%) patients had COPD (black n=182; white n= 802) at the timeof CABG (N =4,801).The median follow-up for study participants was 4.4 years. COPD wasobserved to be a statistically significant predictor of decreased survival independent of racefollowing CABG (no COPD: HR=1.0; white COPD: adjusted HR=1.9, 95% CI=1.7–2.3;black COPD: adjusted HR=1.6, 95% CI=1.1–2.2).

Conclusion: Contrary to the expected increased risk of mortality among black COPDpatients in the general population, a similar survival disadvantage was not observed inour CABG population.

Keywords: COPD, CABG, survival, long-term, race

INTRODUCTIONChronic obstructive pulmonary disease (COPD) affects an esti-mated 15 million Americans and is an important predictor ofmortality following coronary artery bypass grafting (CABG) (1–4). The prevalence of COPD among CABG patients varies from11 to 25.8% (1, 2, 4). COPD recently has become the third leadingcause of death in the United States (5). Risk factors potentiallyassociated with cardiovascular disease and COPD include historyof smoking, increasing age, exposure to air pollution, and lowersocioeconomic position, with smoking being the lead risk factor(6, 7). Between 1980 and 2000, COPD death rates increased by67% in whites and 87% in blacks even though COPD is moreprevalent in whites (6).

Survival paradoxes are well-documented in the cardiovascularliterature. Conventional cardiovascular risk factors such as blackrace, hypercholesterolemia, hypertension, and obesity are asso-ciated with increased survival among some patient populations(8, 9). For example, a recent examination of the national Soci-ety of Thoracic Surgeons (STS) Adult Cardiac Surgery Databasereported a survival advantage among obese patients after CABGcompared with non-obese patients (9).

The rationale for the current study was to determine if sim-ilar reverse epidemiologic findings are observed among blackCOPD patients undergoing isolated CABG in our rural, racially

dichotomous population. Consistent with the expected forceof mortality among blacks, we hypothesized that white COPDpatients would have better long-term survival than black COPDpatients (6).

MATERIALS AND METHODSSTUDY DESIGNThis was a retrospective cohort study of patients undergoing first-time, isolated CABG at the East Carolina Heart Institute between2002 and 2011. Demographic data, comorbid conditions, coro-nary artery disease (CAD) severity, and surgical data were collectedat the time of surgery. Patients with COPD were compared withthose without COPD. Only black and white patients were includedto minimize the potential for residual confounding (∼1% otherraces). Racial identity was self-reported. Emergent cases were con-sidered a clinically different population with a different etiologyfollowing surgery and were excluded in our analysis (n= 105).The study was approved by the Institutional Review Board at theBrody School of Medicine, East Carolina University.

DEFINITIONSCOPD was classified based on severity using the following criteriaat the time of surgery: Mild: FEV1 60–75% of predicted, and/or onchronic inhaled or oral bronchodilator therapy; Moderate: FEV1

www.frontiersin.org April 2013 | Volume 1 | Article 4 | 1

Efird et al. COPD and survival after CABG

50–59% of predicted, and/or on chronic steroid therapy aimedat lung disease; Severe: FEV1 < 50% predicted, and/or room airpO2 < 60 or room air pCO2 > 50. Mortality was defined as anycause of death postoperatively. CAD was defined as at least 50%stenosis and confirmed by angiography before surgery.

SETTINGThe East Carolina Heart Institute is a 120-bed cardiovascularhospital located in the center of eastern North Carolina, a ruralregion with a large black population. Cardiovascular disease is thenumber one cause of death in North Carolina with an unequalburden occurring in eastern North Carolina (10). The instituteis a population-based tertiary referral center. Nearly all patientstreated at the East Carolina Heart Institute live and remain withina 150 mile radius of the medical center.

DATA COLLECTION AND FOLLOW-UPThe primary sources of data extraction were the STS Adult CardiacSurgery Database and the electronic medical record at the BrodySchool of Medicine.

Cardiovascular surgery information at our facility has beenreported to the STS since 1989. Data quality and cross-field valida-tion are routinely performed by the Epidemiology and OutcomesResearch Unit at the East Carolina Heart Institute. An electronicmedical record was introduced at the Brody School of Medicinein 1997. Local and regional clinics were consolidated under a sin-gle electronic medical record in 2005 which allowed for efficientpatient follow-up. The electronic medical record system appliesmultiple logic comparisons to reliably reduce mismatching ofpatient data across clinics and follow-up visits. The STS databaseis linked to the electronic medical record through a unique patientmedical record number. COPD status was not collected in ourdatabase prior to 2002.

The National Death Index was used to obtain death dates forpatients lost to follow-up and also used to validate death informa-tion captured in our electronic medical record (11–13). Linkagewith the National Death Index was based on a multiple criteria,deterministic matching algorithm (13). In our database, less than5% of validated deaths failed to correctly match with the NationalDeath Index.

STATISTICAL ANALYSISCategorical variables were reported as frequency and percent-age while continuous variables were reported as mean± standarddeviation, median, and range. Variables not previously categorizedwere divided into quartiles prior to statistical analysis. Quartilecategorization is advantageous because it limits the influence ofoutliers and allows for the assessment of trend across categories.Follow-up time was measured from the date of surgery to thedate of death or censoring. Survival probabilities were computedusing the Kaplan–Meier product limit method and stratified byCOPD and race. The log-rank test was used to compare sur-vival between patients with and without COPD and among COPDpatients by race. Cox proportional hazard regression models wereused to compute hazard ratios (HR) and 95% confidence inter-vals (CI) for long-term mortality. The initial multivariable modelsincluded variables that have been previously reported to be asso-ciated with cardiovascular-related mortality, regardless of their

statistical significance in our dataset. These included age, sex,race, hypertension, CAD severity, congestive heart failure (CHF),and prior stroke. The post hoc addition of other variables intothe model was performed in a pairwise fashion. The test statisticof Grambsch and Therneau was used to check the proportionalhazards assumption (14). Statistical significance for categoricalvariables was tested using the chi-square (χ2) method and theKruskal–Wallis procedure for continuous variables. PTrend wascomputed using a likelihood ratio test. Temporality during thestudy period was assessed by stratifying the analysis by two timeperiods.

Few values were missing (<1% for included variables). How-ever, when values were missing they were entered into the regres-sion models as a separate category. A sensitivity analysis withmissing values excluded also was performed to confirm thatmodel beta coefficients did not substantively differ from the aboveresults.

Statistical significance was defined as p < 0.05. SAS Version 9.3(Cary, NC, USA) was used for all analyses.

RESULTSA total of 984 (20%) patients had COPD (black n= 182; whiten= 802) at the time of CABG (N = 4,801). The prevalence ofCOPD was higher in white vs. black patients (21 vs. 18%)(p < 0.05). The severity of COPD among black vs. white patients,defined as none, mild, moderate/severe, was not statistically dif-ferent (p= 0.10). Patient characteristics are described in Table 1.The median follow-up for study participants was 4.4 years.

COPD patients had a lower body mass index (BMI)(mean= 29± 6.0 vs. 30± 5.5; p < 0.01) and were more likely to bewhite (82 vs. 78%; p < 0.05) than non-COPD patients. On admis-sion, COPD patients were more likely to have left main disease(31 vs. 25%), hypertension (86 vs. 82%), CHF (35 vs. 23%), renalfailure (4 vs. 2%), peripheral arterial disease (PAD) (26 vs. 13%),previous myocardial infarction (MI) (55 vs. 47%), and previousstroke (11 vs. 8%) (p < 0.01). COPD patients also were more likelyto be recent smokers (55 vs. 26%) and have non-elective CABG(57 vs. 50%) (p < 0.01).

Kaplan–Meier unadjusted survival curves are shown inFigure 1. The 5-year survival for patients with and without COPDwas 73 and 87%, respectively (p < 0.0001). The 5-year survival forblack and white COPD patients was similar (73%).

The unadjusted HR for COPD was 2.1 (95% CI= 1.8–2.4)compared with no COPD. After adjusting for age, sex, race, hyper-tension, CAD severity, CHF, and prior stroke, the HR decreasedto 1.9 (95% CI= 1.6–2.2). COPD was observed to be a statis-tically significant predictor of decreased survival independent ofrace following CABG (no COPD: HR= 1.0; white COPD: adjustedHR= 1.9,95% CI= 1.7–2.3; black COPD: adjusted HR= 1.6,95%CI= 1.1–2.2) (Table 2). The multivariable results did not substan-tively change with the pairwise addition of other variables listedin Table 1.

DISCUSSIONThe significantly increased risk of mortality among COPD patientsafter CABG in our study (adjusted HR= 1.9, 95% CI= 1.6–2.2)is consistent with previous reports (1, 2, 4). A review of 33,137

Frontiers in Public Health | Epidemiology April 2013 | Volume 1 | Article 4 | 2

Efird et al. COPD and survival after CABG

Table 1 | Patients characteristics and univariable survival after CABG (N =4,801).

Characteristic No COPD n (%) COPD Univariable HR (95% CI)

Black n (%) White n (%)

Overall 3,817 (80) 182 (4) 802 (16) 2.1 (1.8–2.4)§

Age (years)

Q1 (≤56) 987 (26) 53 (29) 178 (22) 1.0 Referent

Q2 (>56–64) 995 (26) 43 (24) 223 (28) 1.5 (1.1–1.9)

Q3 (>64–72) 998 (26) 46 (25) 218 (27) 2.3 (1.9–2.9)

Q4 (>72) 837 (22) 40 (22) 183 (23) 3.9 (3.2–4.9)

Mean±SD, median (range) 64±10, 64 (26–90) 63±10, 63 (33–85) 64±10, 65 (31–89) PTrend < 0.0001

Sex

Male 2,723 (71) 117 (64) 595 (74) 1.0 Referent

Female 1,094 (29) 65 (36) 207 (26)†

1.3 (1.1–1.5)

Race

White 2,967 (78) – – 1.0 Referent

Black 850 (22) – – 1.2 (1.03–1.4)

BMI (kg/m2)*

Obese (≥30) 1,716 (45) 78 (43) 314 (39) 1.0 Referent

Overweight (25–29.9) 1,461 (38) 64 (35) 268 (33) 1.3 (1.1–1.5)

Normal (18.5–24.9) 622 (16) 34 (19) 204 (25) 1.8 (1.5–2.2)

Underweight (<18.5) 18 (<1%) 6 (3) 16 (2)††

2.5 (1.4–4.4)

Mean±SD, median (range) 30±5.7, 29 (13–66) 29±6.3, 29 (17–53) 29±6.0, 28 (15–51)††

PTrend < 0.0001

Status

Elective 1,908 (50) 73 (40) 351 (44) 1.0 Referent

Non-elective 1,908 (50) 109 (60) 451 (56)††

1.2 (1.04–1.4)

CAD severity

1 Vessel 268 (7) 10 (5) 41 (5) 1.0 Referent

2 Vessel 971 (25) 41 (23) 218 (27) 1.6 (1.1–2.4)

3 Vessel 2,578 (68) 131 (72) 543 (68) 1.9 (1.4–2.8)

PTrend < 0.0001

Left main disease

No 2,870 (75) 127 (70) 555 (69) 1.0 Referent

Yes 947 (25) 55 (30) 247 (31)††

1.1 (0.93–1.3)

Recent smoker

No 2834 (74) 85 (47) 362 (45) 1.0 Referent

Yes 983 (26) 97 (53) 440 (55)††

0.98 (0.85–1.1)

Hypertension

No 680 (18) 16 (9) 122 (15) 1.0 Referent

Yes 3,137 (82) 166 (91) 680 (85)††

1.4 (1.2–1.7)

Diabetes

No 2,309 (60) 94 (52) 490 (61) 1.0 Referent

Yes 1,508 (40) 88 (48) 312 (39) 1.4 (1.3–1.7)

Congestive heart failure

No 2,927 (77) 107 (59) 532 (66) 1.0 Referent

Yes 890 (23) 75 (41) 270 (34)††

2.1 (1.9–2.5)

Renal failure

No 3,744 (98) 175 (96) 771 (96) 1.0 Referent

Yes 73 (2) 7 (4) 31 (4)††

4.2 (3.2–5.6)

Dialysis

No 3,730 (98) 165 (91) 793 (99) 1.0 Referent

Yes 87 (2) 17 (9) 9 (1)††

4.9 (3.7–6.4)

(Continued)

www.frontiersin.org April 2013 | Volume 1 | Article 4 | 3

Efird et al. COPD and survival after CABG

Table 1 | Continued

Characteristic No COPD n (%) COPD Univariable HR (95% CI)

Black n (%) White n (%)

Peripheral arterial disease

No 3,320 (87) 136 (75) 590 (74) 1.0 Referent

Yes 497 (13) 46 (25) 212 (26)††

2.1 (1.8–2.5)

Prior MI

No 2,007 (53) 77 (42) 367 (46) 1.0 Referent

Yes 1,810 (47) 105 (58) 435 (54)††

1.5 (1.3–1.7)

Prior stroke

No 3,509 (92) 157 (86) 719 (90) 1.0 Referent

Yes 308 (8) 25 (14) 83 (10)††

1.8 (1.5–2.3)

Prior PCI

No 2,876 (75) 135 (74) 619 (77) 1.0 Referent

Yes 941 (25) 47 (26) 183 (23) 0.83 (0.70–0.98)

†p < 0.05; ††p < 0.01; χ2 (Categorical variables), Kruskal–WallisTest (Continuous variables). *Missing category not shown. §COPD vs. no COPD. BMI, body mass index;

CABG, coronary artery bypass grafting; CAD, coronary artery disease; CI, confidence interval; COPD, chronic obstructive pulmonary disease; HR, hazard ratio; MI,

myocardial infarction; PCI, percutaneous coronary intervention; Q1, quartile 1; Q2, quartile 2; Q3, quartile 3; Q4, quartile 4.

FIGURE 1 | CABG, coronary artery bypass grafting; COPD, chronic obstructive pulmonary disease.

CABG cases from hospitals in northern New England foundCOPD to be an independent predictor of decreased long-termsurvival (HR= 1.8, 95% CI= 1.6–2.1) (1). A covariate-matchedanalysis of 3,760 patients in New York that underwent isolatedCABG similarly found an increased risk in long-term mortal-ity for patients with COPD (HR= 1.28, 95% CI= 1.11–1.47)(2). Furthermore, a review of 13,337 CABG procedures in theUnited Kingdom found COPD to be a predictor of long-termsurvival and differences were noted between moderate (HR= 1.3,95% CI= 1.1–1.5) and severe (HR= 1.9, 95% CI= 1.5–2.4) cases(4). Our results also are comparable with population-based datashowing that COPD is a predictor of mortality in the gen-eral population and that COPD is more prevalent in whitepatients (6, 15).

Death rates have been reported to be higher among black vs.white COPD patients, reflecting a greater force of mortality among

blacks in the general population (6, 16). A recent examinationof the national lung transplant list observed that black COPDpatients died sooner than whites; however, the results were con-founded by patients who were removed from the transplant list(17). In contrast, the HR for the black COPD arm in our studywas in the reverse direction, being less than white COPD. In acomparable analysis of patients without COPD, an increased forceof mortality was observed among black PAD patients (no PAD,HR= 1.0; white PAD: adjusted HR= 1.4, 95% CI= 1.1–1.8; blackPAD: adjusted HR= 2.1, 95% CI= 1.5–3.0). The latter highlightsa clear force of mortality among black patients for another chronicdisease population at-risk for cardiovascular disease and empha-sizes the unexpected finding among black COPD patients in ourstudy.

Survival paradoxes in black patients are well-documented inthe literature (18–21). Our data showed no survival advantage

Frontiers in Public Health | Epidemiology April 2013 | Volume 1 | Article 4 | 4

Efird et al. COPD and survival after CABG

Table 2 | Multivariable survival after CABG.

Characteristic Multivariable HR (95% CI)

Main effect

No COPD 1.0 Referent

Black COPD 1.6 (1.1–2.2)

White COPD 1.9 (1.7–2.3)

Age (years)

Q1 (≤56) 1.0 Referent

Q2 (>56–64) 1.4 (1.1–1.8)

Q3 (>64–72) 2.2 (1.7–2.7)

Q4 (>72) 3.5 (2.8–4.3)

PTrend < 0.0001

Sex

Male 1.0 Referent

Female 1.1 (0.92–1.2)

CAD severity

1 Vessel 1.0 Referent

2 Vessel 1.4 (0.97–2.1)

3 Vessel 1.5 (1.1–2.2)

PTrend=0.017

Hypertension

No 1.0 Referent

Yes 1.2 (1.02–1.5)

Congestive heart failure

No 1.0 Referent

Yes 1.8 (1.6–2.1)

Prior stroke

No 1.0 Referent

Yes 1.5 (1.3–1.9)

CABG, coronary artery bypass grafting; CAD, coronary artery disease; CI, con-

fidence interval; COPD, chronic obstructive pulmonary disease; HR, hazard

ratio.

for white COPD patients and a survival paradox for blackCOPD patients cannot be ruled out. Contrary to our hypoth-esis, a statistically significant survival disadvantage was notobserved among black patients in our CABG population. How-ever, uncertainty remains due to the limited number of patientsat-risk toward the end of the study. A possible explanationfor our discordant findings may involve differences in lungfunction and susceptibility to tobacco smoke between blackand white COPD patients (22, 23). Also, differences in theresponse to standard COPD therapies could exist betweenraces (24).

STRENGTHS AND LIMITATIONSOur study is strengthened by its comparatively large sample sizeand long-term follow-up. To the best of our knowledge, this isthe largest study to date examining survival differences amongCOPD CABG patients by race. Furthermore, we were able toaccurately determine time of death using a combination of theNational Death Index and our comprehensive electronic medicalrecord.

Another strength of this study is its target base. A large pri-ority population in eastern North Carolina allowed for us toreport on a group that has experienced historic differences insocioeconomic position and discrimination. Twenty-eight (97%)of the 29 counties in eastern North Carolina fall below the nationalper capita income of $27,915, with half reporting a value less than$20,000 (25). Similarly, 90% of the counties have a higher percent-age of blacks than the national value of 13.1% (25). Our resultsare generalizable to other low-income, rural, and racially diversepopulations.

Pulmonary function tests were not repeated prior to surgery.Accordingly, there may have been misclassification of COPD sta-tus between patients. We were unable to stratify our analyses byCOPD severity and race because of a limited number of patients.Survival after CABG among COPD patients has been shown tovary by disease severity and racial differences could exist acrossthe COPD severity spectrum (4).

Socioeconomic position, education, and income were not col-lected and these factors may have influenced survival (26). Payorstatus, which has been shown in some studies to predict survivalindependent of race, was not consistently collected and conse-quently was not used in our analysis (27). Additionally, we wereunable to reliably estimate socioeconomic position using zip codesbecause a large percentage of patients in our catchment area livein rural areas with postal box addresses.

Patients in this study were recruited over a relatively longperiod (10 years), over which practice methods and clinical caremay have changed considerably. However, results were consis-tent throughout the study after stratifying by two time periods,indicating the robustness of the data to temporal changes. Thestatus of several variables in our analysis may have changedover time. We did not adjust for these variables in a time-dependent manner due to their potential to be in the causalpathway. Similarly, surgical complications and medication usewere not included in our analysis because of their time-dependentstatus.

Cause of death is not recorded in the National Death Index andCOPD status may have been unrelated to their mortality. Althoughwe adjusted for known clinically relevant variables, we acknowl-edge that other unmeasured factors could have influenced ourresults due to the retrospective nature of this study. Retrospec-tive studies also are susceptible to recall and selection bias. Wecannot rule out that the association between COPD and poorsurvival is non-causal in nature and may be due to non-cardiaccauses unassociated with comorbidities or surgical proceduressuch as exacerbations requiring hospitalizations (15). Addition-ally, examining the right tail of the Kaplan–Meier curve (Figure 1),it is observed that the number of individuals at-risk reduces toa few patients after year 6. Probability estimates at these latertime points may not be reliable and must be interpreted withcaution.

We considered missing values to be a distinct category andthey were entered into the regression models as a separate cat-egory rather than being imputed. Imputation methods requiredata to be “missing at random” which is difficult to verifygiven the sparseness and unknown distribution of the missing

www.frontiersin.org April 2013 | Volume 1 | Article 4 | 5

Efird et al. COPD and survival after CABG

values (28, 29). We cannot rule out misclassification bias dueto grouping missing values into a distinct category althoughsuch bias likely is trivial given the small number of miss-ing values. Furthermore, we performed a complete case analy-sis with missing values removed and this did not substantivelyalter our results. We opted not to report the complete analy-sis because this would have reduced power for our multi-variable analysis. Removing values only from the multivariateanalysis also would have resulted in a different dataset beinganalyzed.

Our use of quartile boundaries, while desirable for minimizingthe influence of outliers, may have yielded overly broad cate-gories and the potential for residual confounding. However, thesubstitution of continuous variables in our models did not mate-rially alter results. Except for race, we did not examine interac-tions among other clinically relevant variables included in ourdataset. Given the large number of potential multi-level interac-tions involving the independent variables in our analyses, it isdifficult to interpret such effects. Furthermore, we did not useregression-based tests for interaction because they are knownto have weak power and often fail to detect interactions whenthey exist (30). Multivariable Cox regression models, rather thanpropensity score matching, were used to control for confounding

because of potential “non-collapsibility bias” inherent to logisticregression-based propensity scores and the possible loss of powerdue to incomplete matching (31). Alternative methods such asmachine learning (e.g., random forest algorithm) may introducemisspecification into the propensity score model due to the “blackbox” nature of the algorithm that obscures the etiologic relation-ship between predictors and outcome and were not used in thecurrent analysis (32, 33).

CONCLUSIONContrary to the expected increased risk of mortality among blackCOPD patients in the general population, a similar survival disad-vantage was not observed in the current study. Future research isneeded to confirm our initial findings. Exploring the associationbetween preoperative conditions such as COPD and outcomes,especially in the context of racial disparities in rural populations,may help guide future decisions regarding appropriate clinicalfollow-up and surgical referral.

ACKNOWLEDGMENTSDr. Efird is supported by NIH Grant 5R34DE022272-02. Theauthors would like to thank the East Carolina Heart Institute forproviding resources to conduct this study.

REFERENCES1. Leavitt BJ, Ross CS, Spence B,

Surgenor SD, Olmstead EM,Clough RA, et al. Long-termsurvival of patients with chronicobstructive pulmonary diseaseundergoing coronary arterybypass surgery. Circulation (2006)114:I430–34.

2. Angouras DC, Anagnostopou-los CE, Chamogeorgakis TP,Rokkas CK, Swistel DG, Con-nery CP, et al. Postoperative andlong-term outcome of patientswith chronic obstructive pul-monary disease undergoing coro-nary artery bypass grafting. AnnThorac Surg (2010) 89:1112–8.

3. Centers for Disease and Preven-tion. Chronic obstructive pul-monary disease among adults–United States, 2011. MMWR MorbMortal Wkly Rep (2012) 61:938–43.

4. O’Boyle F, Mediratta N, ChalmersJ, Al-Rawi O, Mohan K, ShawM, et al. Long-term survivalof patients with pulmonary dis-ease undergoing coronary arterybypass surgery. Eur J CardiothoracSurg (2012) 43:697–703.

5. Minino AM, Murphy SL, Xu J,Kochanek KD. Deaths: final datafor 2008. Natl Vital Stat Rep (2011)59:1–126.

6. Mannino DM, Homa DM, Akin-bami LJ, Ford ES, Redd SC.Chronic obstructive pulmonarydisease surveillance – United

States, 1971–2000. MMWR Sur-veill Summ (2002) 51:1–16.

7. Gan WQ, Fitzgerald JM, Carl-sten C, Sadatsafavi M, BrauerM. Associations of ambientair pollution with chronicobstructive pulmonary diseasehospitalization and mortal-ity. Am J Respir Crit Care Med(2013). doi:10.1164/rccm.201211-2004OC

8. Kalantar-Zadeh K, Kovesdy CP,Derose SF, Horwich TB, FonarowGC. Racial and survival para-doxes in chronic kidney disease.Nat Clin Pract Nephrol (2007) 3:493–506.

9. Shahian DM, O’brien SM, ShengS, Grover FL, Mayer JE, JacobsJP, et al. Predictors of long-term survival after coronaryartery bypass grafting surgery:results from the Society of Tho-racic Surgeons Adult CardiacSurgery Database (the ASCERTstudy). Circulation (2012) 125:1491–1500.

10. Morris PJ. Heart disease andstroke in North Carolina. N C MedJ (2012) 73:448–9.

11. Lash TL, Silliman RA. A com-parison of the National DeathIndex and Social Security Admin-istration databases to ascertainvital status. Epidemiology (2001)12:259–61.

12. Welke KF, Ferguson TB Jr.,Coombs LP, Dokholyan RS, Mur-ray CJ, Schrader MA, et al. Validity

of the Society of Thoracic Sur-geons National Adult CardiacSurgery Database. Ann ThoracSurg (2004) 77:1137–39.

13. Morales DL, McClellan AJ, JacobsJP. Empowering a database withnational long-term data aboutmortality: the use of nationaldeath registries. Cardiol Young(2008) 18(Suppl 2):188–95.

14. Grambsch P, Therneau T. Propor-tional hazards tests and diagnos-tics based on weighted residuals.Biometrika (1994) 81:515–26.

15. Garcia-Aymerich J, Serra Pons I,Mannino DM, Maas AK, MillerDP, Davis KJ. Lung functionimpairment, COPD hospitalisa-tions and subsequent mortality.Thorax (2011) 66:585–90.

16. Bharmal N, Tseng CH, KaplanR, Wong MD. State-level varia-tions in racial disparities in lifeexpectancy. Health Serv Res (2012)47:544–55.

17. Lederer DJ, Benn EK, Barr RG,Wilt JS, Reilly G, Sonett JR, etal. Racial differences in waitinglist outcomes in chronic obstruc-tive pulmonary disease. Am JRespir Crit Care Med (2008) 177:450–54.

18. Bloembergen WE, Port FK,Mauger EA, Wolfe RA. Causes ofdeath in dialysis patients – racialand gender differences. J Am SocNephrol (1994) 5:1231–42.

19. Owen WF, Chertow GM, LazarusJM, Lowrie EG. Dose of

hemodialysis and survival –differences by race and sex. JAMA(1998) 280:1764–8.

20. Agodoa L, Eggers P. Racial andethnic disparities in end-stage kid-ney failure – survival paradoxesin African-Americans. Semin Dial(2007) 20:577–85.

21. Kucirka LM, Grams ME, LesslerJ, Hall EC, James N, Massie AB,et al. Association of race andage with survival among patientsundergoing dialysis. JAMA (2011)306:620–26.

22. Harik-Khan RI, Fleg JL, MullerDC, Wise RA. The effect ofanthropometric and socioeco-nomic factors on the racial dif-ference in lung function. Am JRespir Crit Care Med (2001) 164:1647–54.

23. Chatila WM, Wynkoop WA,VanceG, Criner GJ. Smoking pat-terns in African Americans andwhites with advanced COPD.Chest (2004) 125:15–21.

24. Kirkpatrick DP, Dransfield MT.Racial and sex differences inchronic obstructive pulmonarydisease susceptibility, diagnosis,and treatment. Curr Opin PulmMed (2009) 15:100–104.

25. United States Census Bureau.2010. US Census Bureau State &County Quick Facts. Washington:U.S. Government Printing Office[Online; cited 2013 March 10].Available from: http://quickfacts.census.gov/qfd/index.html

Frontiers in Public Health | Epidemiology April 2013 | Volume 1 | Article 4 | 6

Efird et al. COPD and survival after CABG

26. Koch CG, Li L, Kaplan GA,Wachterman J, Shishehbor MH,Sabik J, et al. Socioeconomic posi-tion, not race, is linked to deathafter cardiac surgery. Circ Cardio-vasc Qual Outcomes (2010) 3:267–76.

27. Zacharias A, Schwann TA, Rior-dan CJ, Durham SJ, Shah A, HabibRH. Operative and late coronaryartery bypass grafting outcomesin matched African-Americanversus Caucasian patients: evi-dence of a late survival-Medicaidassociation. J Am Coll Cardiol(2005) 46:1526–35.

28. Schafer JL, Graham JW. Missingdata: our view of the state ofthe art. Psychol Methods (2002)7:147–77.

29. Heitjan DF. Incomplete data: whatyou don’t know might hurt you.Cancer Epidemiol Biomarkers Prev(2011) 20:1567–70.

30. Aiken LS, West SG, Reno RR. Mul-tiple regression: testing and inter-preting interactions. In MultipleRegression: Testing and Interpret-ing Interactions. Thousand Oaks:SAGE Publications (1991). 156 p.

31. Efird JT, Lea S, Toland A, PhillipsCJ. Informational odds ratio:a useful measure of epidemio-logic association in environmentexposure studies. Environ HealthInsights (2012) 6:17–25.

32. Breiman L. Random forests. MachLearn (2001) 45:5–32.

33. Lee BK, Lessler J, Stuart EA.Improving propensity score

weighting using machine learning.Stat Med (2010) 29:337–46.

Conflict of Interest Statement:The authors declare that theresearch was conducted in theabsence of any commercial orfinancial relationships that couldbe construed as a potential conflictof interest.

Received: 15 February 2013;accepted: 16 March 2013; publishedonline: 15 April 2013.Citation: Efird JT, O’Neal WT,Anderson CA, O’Neal JB, KindellLC, Ferguson TB, ChitwoodWR and Kypson AP (2013)The effect of race and chronicobstructive pulmonary disease

on long-term survival aftercoronary artery bypass grafting.Front. Public Health 1:4. doi:10.3389/fpubh.2013.00004This article was submitted to Fron-tiers in Epidemiology, a specialty ofFrontiers in Public Health.Copyright © 2013 Efird, O’Neal,Anderson, O’Neal, Kindell, Fer-guson, Chitwood and Kypson.This is an open-access articledistributed under the terms ofthe Creative Commons Attribu-tion License, which permits use,distribution and reproduction inother forums, provided the origi-nal authors and source are cred-ited and subject to any copyrightnotices concerning any third-partygraphics etc.

www.frontiersin.org April 2013 | Volume 1 | Article 4 | 7